AI trade approval gates: Why Rafael Would Have Rejected the Leveraged Position

An unfamiliar leveraged position means your software had authority you should not have granted. A final approval gate keeps an AI-generated order from becoming a live trade until you review the current price, size, stop, and risk.

Consider Rafael, an illustrative composite: at 6:17 a.m. in his Lisbon kitchen, he opened his exchange app while the kettle clicked off. A position he did not recognize sat at the top of the screen, larger than his usual trades and using leverage he had intended to avoid.

The chart had moved against it overnight. Rafael had given a bot permission to trade after weeks of watching its alerts. He assumed the settings would keep it within his limits. Now closing the position would lock in a loss, while holding it meant accepting more risk than he had chosen when he went to bed. The bad ending was clear: a single unattended trade could do more damage than the series of small, planned losses he had learned to tolerate.

The failure began before the order was placed

The software may have followed its rules exactly. That does not make the outcome acceptable.

An automated strategy can see a setup, calculate a position size, and send an order faster than a person. It cannot know whether you have changed your risk limit after a difficult week, whether an open position already uses your available exposure, or whether you simply do not want leverage on that market today.

Those decisions are part of trading discipline. They belong to the person whose capital is at risk.

The real failure was the permission model. Rafael had treated a trade signal as a trade decision, then allowed the system to execute the decision without asking again. The order was unfamiliar because the final judgment had been removed from the process.

That gap gets wider in stressed markets, when prices move quickly and assumptions built into a strategy can break. Pre-trade controls, post-trade checks, monitoring, and governance all exist for the same reason: systems need boundaries when conditions change.

A signal can be useful without receiving execution authority

A trading assistant can still do meaningful work before the decision reaches you. It can identify a setup, prepare an order, show the proposed entry and stop, and calculate what the loss could be if the stop is hit.

Then it should wait.

That pause changes the practical question from “Did the system find a trade?” to “Does this trade fit my plan right now?” You can reject an order because the position size exceeds your limit, because the stop is too wide, because you already have correlated exposure, or because the market has moved beyond the entry you would accept.

Those are not failures of automation. They are the reason the approval step exists.

TraderCoach is built around that boundary. The AI can generate and queue trade signals, but a human approves or rejects each one before anything executes. The queue gives you a visible decision point while the trade is still optional.

For a closer look at how risk can change while an order waits, see The Friday Order That Needed a Second Decision, Before Risk Changed.

The approval screen should make risk visible

An approval gate only helps when the information needed for a decision is clear. “Buy” or “sell” alone tells you almost nothing.

Before approving a queued order, check:

• The proposed position size and the dollar amount at risk if the stop is reached.
• The entry price, stop level, and whether the market has already moved away from the original setup.
• Your existing positions, especially holdings that could move together.
• Any leverage attached to the order and whether it matches your written limits.
• The maximum loss you are willing to accept on this trade and across the account.

Illustration: if a proposed position risks $145 and your fixed limit is $100, the correct next step is not to admire the setup. It is to reduce the size, change the plan, or reject the order. Mara’s $145 loss risk. Her $100 limit holds. shows why that comparison needs to happen before execution.

The same review protects lower-experience traders from a common mistake: treating a confident-looking signal as proof. A signal is an input. Your risk limit is the constraint.

The position Rafael would have rejected

Rafael did not need a better explanation after the fact. He needed a moment before the order went live.

With an approval gate, the leveraged position would have appeared in a queue. He could have seen the size before it opened, compared it with his limit, and rejected it from the kitchen table. The trade might still have been a valid setup on paper. It did not fit the way he wanted to trade.

The morning after that decision looks quieter. There is no unknown position to unwind, no attempt to reconstruct which setting allowed it, and no pressure to make a rushed choice because a bot made the first one.

Keep the strategy accountable to your rules. Keep the final decision with the person who has to live with the result.